Amazon Weighs US$8 Billion Nvidia Chip Sale To Investors - BusinessToday Malaysia

This is not an asset liquidation — it is a structural pivot in hyperscaler compute strategy. Amazon pushing $8B of Nvidia GPUs into the secondary market signals one thing: Trainium's custom silicon has crossed the production threshold where marginal AI workloads no longer require H100/B200 capacity. The compute land-grab era of 2023-24 is yielding to a compute-efficiency era. For Nvidia, this creates a structural paradox. Its largest enterprise customers are simultaneously its volume backbone and its exit risk. An $8B offload implies Amazon's GPU fleet has reached saturation relative to actual training throughput — or more precisely, Trainium is absorbing the workload that previously demanded Nvidia silicon. This is not demand contraction; it is demand migration. Competitive ripple: Microsoft and Google remain deeply CUDA-locked and face a strategic dilemma. Oracle's AI cloud becomes significantly more attractive if it can offer Nvidia-native environments at competitive pricing, directly siphoning enterprise workloads. The chip is transitioning from strategic commodity to tradable asset class — the same market-formation logic as the used-server market in 2015. 12-24 month outlook: Expect two to three additional hyperscaler GPU offloading events. Nvidia's data center revenue growth will decouple from pure unit volume, pivoting toward software ecosystem lock-in. The real moat is no longer the silicon. It is the gravitational pull of the ecosystem.